Teacher Forcing

Teacher forcing is a strategy for training rnn networks that uses ground truth as input, instead of model output from a prior time step as an input.

 

But, the approach can also result in models that may be fragile to test data.

 

Curriculum learning, a variation of forced learning, involves randomly choosing to use the ground truth output or the generated output form the previous time step as input for the current time step. 

 

 

 

참고한 사이트 링크: https://machinelearningmastery.com/teacher-forcing-for-recurrent-neural-networks/